Current status
Deferred goal-anchor. This issue is no longer a v13.2 release-core container, and it is not implementation authority for immediate file moves or architecture rewrites.
Roadmap re-triage is the purpose of this edit: ROADMAP.md carries #14304 with zero release weight until the stale body is re-triaged. Mnemosyne's 2026-07-04 re-triage identified the mismatch, and Grace accepted disposition (a): re-scope in place. This body is the mechanical re-scope under that author authorization.
Goal
Keep Agent OS architecture quality from drifting as the organism grows: Brain domain placement must be legible, Body and Brain idioms must stay distinct, and any future architecture-enforcement work must be grounded in a version-bound intended architecture SSOT instead of a stale release-epic body.
This issue should act as the parked goal anchor. If the work re-enters an active roadmap slot, it scopes into self-selected lane epics; peers claim lanes themselves. It is not a parent-of-sub-lanes and should not accumulate implementation leaves directly.
Re-scoped kernel
- Brain domain mapping: the
ai/ tree is still partly organized by technical layer instead of domain ownership. The durable concern is domain-first Brain placement with thin technical edges, not a big-bang folder move.
- Architecture authority: the intended physical architecture SSOT must cohere with the conceptual Body / Brain / Neural Link authority tracked by #13846. A Decision Record is required before enforcement hardens.
- Coupling evidence: memory-graph coupling, graph/dream placement, immune/diagnostics placement, and orchestrator ownership need live import/call evidence before any boundary claim becomes implementation authority.
- Migration mechanics: any future movement needs codemod/import rewriting, mirrored test movement, script/config/path sweeps, docs/ADR path updates, and snapshot protection. No merge-storm.
- Body-idiom integrity: #14714 and the Base/Neo best-practices audits showed the architecture-quality concern is broader than
ai/ folders. Body code written in Brain style is now an explicit strand of this goal.
- Corrected cruft claims:
ai/data/ was not safe delete work; the lazy-edges queue needed re-home first. ai/docs/ was not public docs cruft; it is internal Agent OS process substrate and needs registered relocation if moved.
Deferred gates
No implementation lane should proceed from this issue until these gates are version-bound:
- Lane-0 intended architecture SSOT exists and is reconciled with #13846.
- Coupling V-B-A matrix exists for the disputed boundaries.
- Migration tooling exists for imports, mirrored tests, scripts/configs, and docs/ADR path references.
- The roadmap gives this goal an active slot again; while v13.2 prioritizes the five current cornerstones, this remains deferred.
- Enforcement is staged: advisory during migration, then rule-by-rule blocking with explicit exception TTLs.
Current disposition
Deferred by design; not code-ready. Keep open as the durable architecture-quality goal anchor, but do not pick implementation work directly from this issue body.
Source anchors
- Re-triage verdict: #14304 comment by Mnemosyne on 2026-07-04.
- Author disposition: #14304 comment by Grace on 2026-07-04 accepting re-scope.
- Settled model: #14304 comment by Grace on 2026-06-28: goal-anchor, not parent-of-sub-lanes.
- Related source artifacts: #13846, #14302, #14305, #14306, #14300, #14307, #14714.
Current status
Deferred goal-anchor. This issue is no longer a v13.2 release-core container, and it is not implementation authority for immediate file moves or architecture rewrites.
Roadmap re-triage is the purpose of this edit:
ROADMAP.mdcarries #14304 with zero release weight until the stale body is re-triaged. Mnemosyne's 2026-07-04 re-triage identified the mismatch, and Grace accepted disposition (a): re-scope in place. This body is the mechanical re-scope under that author authorization.Goal
Keep Agent OS architecture quality from drifting as the organism grows: Brain domain placement must be legible, Body and Brain idioms must stay distinct, and any future architecture-enforcement work must be grounded in a version-bound intended architecture SSOT instead of a stale release-epic body.
This issue should act as the parked goal anchor. If the work re-enters an active roadmap slot, it scopes into self-selected lane epics; peers claim lanes themselves. It is not a parent-of-sub-lanes and should not accumulate implementation leaves directly.
Re-scoped kernel
ai/tree is still partly organized by technical layer instead of domain ownership. The durable concern is domain-first Brain placement with thin technical edges, not a big-bang folder move.ai/folders. Body code written in Brain style is now an explicit strand of this goal.ai/data/was not safe delete work; the lazy-edges queue needed re-home first.ai/docs/was not public docs cruft; it is internal Agent OS process substrate and needs registered relocation if moved.Deferred gates
No implementation lane should proceed from this issue until these gates are version-bound:
Current disposition
Deferred by design; not code-ready. Keep open as the durable architecture-quality goal anchor, but do not pick implementation work directly from this issue body.
Source anchors